On December 21, 2024, the 2nd International Conference on Artificial Intelligence Generated Content and Large Model Application Innovation Conference (AIGC 2024) was held at the Metropark International Hotel in Beijing. The AIGC International Conference is the world's first academic conference focusing on generative artificial intelligence. The first conference last year attracted great attention from the academic community. In the past year, large language models, Vincentian graphs, Vincentian videos, graphogenic videos, and large model architectures , training data, and the development of multi-modal large models have advanced by leaps and bounds, and the values ​​were corrected on the first day of the meeting It was the first day that OpenAI released a new generation of o3 inference large model. Against this background, the second AIGC was held, attracting top scholars, experts and industry leaders from all over the world to discuss the future direction of artificial intelligence technology and its application. Wide range of applications in various fields.

This conference is hosted by the AIGC 2024 Conference Organizing Committee. Professor Qiu Xipeng from the School of Computer Science at Fudan University served as the chairman of the conference; Professor Ouyang Wanli, a leading scientist from the Shanghai Artificial Intelligence Laboratory, served as the co-chairman of the conference; Professor Zhao Feng from the University of Science and Technology of China and Professor Ou Zhonghong from Beijing University of Posts and Telecommunications co-chaired the conference program committee; Professor Miao Duoqian from Tongji University , Professor Yan Junchi of Shanghai Jiao Tong University and Professor Noor Zaman Jhanjhi of Taylor’s University jointly serve as the chairman of the technical committee.

The conference was hosted by Shanghai Zhongzhan Guorui Exhibition Co., Ltd. and also received support from Peking University, Tsinghua University, Beijing University of Posts and Telecommunications, Xi'an Jiaotong University, Chinese Academy of Sciences, Beijing University of Information Science and Technology, Zhejiang University of Technology, University of Hong Kong, Hong Kong Baptist University, Western Scholars from well-known domestic and foreign universities such as Hu University, Chongqing Normal University, Northeastern University, King's College London, University of Reading, Binghamton University, University of Illinois at Chicago, etc., as well as technical experts from well-known enterprises such as Huawei, Tencent, and Microsoft Health Futures strong support.

In the two-day agenda, the conference includes the main forum (conference report), cutting-edge technology theme forum, embodied intelligence and intelligent interaction theme forum, trusted AI and large model security theme forum, application Innovation theme forum, large model technology exploration and practice theme forum, AI cutting-edge application theme forum, as well as various academic activities and links such as oral reports and poster reports.

The opening ceremony and main forum of the conference were chaired by Professor Ou Zhonghong, Chairman of the Program Committee. Professor Ou pointed out that AIGC has gradually become an important force driving social change from a cutting-edge theory in academia. From high-precision code generation to complex artistic creation, from real-time virtual assistants to digital twins, AIGC is reshaping production methods and industrial ecology at an unprecedented speed.

In this context, as an important academic exchange platform in the field of AIGC, this conference continues the successful experience of the first conference and promotes sustainable innovation of large models and AIGC technology through cross-field thought collision and practice sharing. . Professor Ou hopes that elites from various research teams and industry experts can share with everyone the basic theoretical breakthroughs, cutting-edge algorithm design and practical orientation of large models at AIGC 2024.The applied innovative solutions provide valuable inspiration for a deep understanding of AIGC’s future blueprint. 24 —

Main Forum - Conference Report

Sun Jian, a professor and outstanding student at Xi’an Jiaotong University and deputy dean of the School of Mathematics and Statistics, delivered a speech at the conference entitled "The Mathematics of Generative Artificial Intelligence" and Fundamentals of Statistics" conference report.

Professor Sun Jian first explained the principles behind generative AI from the perspective of mathematical statistics, including basic principles such as probability distribution and sample generation. Then, Professor Sun Jian introduced generative AI in detail The application and future research trends of AI in model training, algorithm optimization, and medical fields.

Professor Sun said: “Generally speaking, generative AI is actually a distribution learning change of this high-dimensional multi-modal complex data, and it actually does distribution learning, so the samples it uses must not comply with the laws of physics. We try to make it consistent with the laws of physics. How to interact with the physical world? How to create more controllable, realistic and generalized artificial intelligence? Yes, there are also various issues such as the generation of dynamic systems, the generation of logic, the generation of mathematical inductive reasoning, large multi-modal models of medical imaging, etc. We must not only care about the engineering implementation, but also pay attention to the behind-the-scenes problems. Mathematically based model

Zhang Yue, a tenured professor at West Lake University, published a paper titled "LLM. Reasoning and Generalization" conference report.

In his speech, Professor Zhang introduced the importance of logical reasoning for general artificial intelligence, and introduced the concept of logical reasoning in detail. Then, Professor Zhang Yue introduced Out-of-Distribution Generalization, thanks to ChatGPT 4 Such large models are only statistical and have no causality, which brings about the problem of poor generalization ability outside the distribution. Professor Zhang Yue introduced two solutions to this problem in his speech. One is through data governance. The other is to change the model's architecture and training methods to improve the causality of the model. Finally, Professor Zhang said in the speech. Jie Zhong said that today’s artificial intelligence lacks rationality and is different from human intelligence, but artificial intelligence and human intelligence also have intersections. If artificial intelligence is to be more trustworthy, it must be closer to human intelligence.

President of Chongqing Normal University, IRSS/I2CICC/CAAI/CCF Professor Wang Guoyin, Fellow, IEEE SM, and Vice Chairman of CAAI, delivered a conference report titled "Brain Cognition Inspired Artificial Intelligence".

Professor Wang Guoyin's report mainly focused on the topic. It unfolds in three aspects: first, the inspiration of brain cognition for artificial intelligence research; second, the limitations of current artificial intelligence inspired by brain cognition; third, the future trends and inspirations of the integration of artificial intelligence and cognitive science.Show. The report reviews the role of cognitive science in the three mainstream academic branches that promote the development of AI, and analyzes the differences and contradictions between the cognitive mechanism of the human brain and the computing mechanism of AI systems. These differences are considered to be the relationship between AI systems and human behavior. One of the reasons for the inconsistency. In his report, Professor Wang summarized eight important research directions inspired by brain cognition, including highly imitating biological information processing, structural and functional balance of large-scale deep learning models, multi-granularity joint problem solving, and AI models that simulate specific brain structures. Co-processing mechanisms that physically separate perception processing and interpretation analysis, embodied intelligence that combines brain cognitive mechanisms with AI computing mechanisms, intelligent simulation from individual intelligence to group intelligence (social intelligence), and AI-assisted brain cognitive intelligence. Professor Wang’s report pointed out that understanding the working mechanism of the human brain is crucial to developing more intelligent, efficient and humane AI systems. Through interdisciplinary research, combining brain science with advanced AI algorithms, it is expected to achieve more interpretable and robust artificial intelligence in the future.

Representing the team of Professor Wang Xingwei, Vice President of Northeastern University, CCF Fellow and Distinguished Young Scholar, Professor Ren Feiliang of Northeastern University delivered a conference report entitled "Thoughts on Several Issues in the Development of Generative Artificial Intelligence".

In his report, Professor Ren Feiliang introduced AIGC’s work from three aspects: whether mankind has entered the era of generative artificial intelligence, thoughts on several issues regarding the development of generative artificial intelligence technology, and thoughts on several issues regarding the future development of generative artificial intelligence. Current situation and existing problems. In terms of technology, Professor Ren Feiliang gave a detailed introduction to the current AIGC technology architecture, task processing mechanism, operating mechanism, macro tasks, micro tasks, privacy and ethics, large models, small models, and model evaluation issues. Regarding the future development of AIGC, Professor Ren first analyzed the core competition and competitive strategies of AIGC. Regarding AIGC research, should we continue to "run blindfolded" or "move forward with our eyes open"?

Should we pay attention to backward compatibility? How to empower intelligent networks? Discuss the issues. Finally, Professor Ren called on higher education to adapt to the development of AIGC. With the current development of AI, scientific researchers need more than ever "organized scientific research, and collaborative innovation of industry, academia and research" more than ever. p>

—AIGC 2024 —

Theme sub-forums

On the afternoon of December 21, the conference set up four theme forums, including "Frontier Technology Theme Forum", "Trusted AI and Large Model Security Theme" Forum", "Embodied Intelligence and Intelligent Interaction Theme Forum" and "Application Innovation Theme Forum". The discussions in these four forums cover generative The multi-dimensional challenges and applications of AI and large model technology profoundly reveal the cutting-edge trends of AI development and its far-reaching impact in different industries. The development of generative AI and large models requires more attention to ethics and controllability while making technological breakthroughs. The fit between nature and application scenarios. Through in-depth exchanges and discussions in the forum, participants discussed the research directions and products of future AI technology.It has a clearer understanding of industrial applications and further promotes interdisciplinary cooperation and innovation.

Frontier Technology Theme Forum

At the Frontier Technology Theme Forum, academic experts conducted in-depth discussions around the latest advances in generative artificial intelligence and deep learning. The forum was hosted by Liu Peng, Assistant Professor of Singapore Management University, and invited heavyweight guests such as Huang Gao, Associate Professor of Tsinghua University, Yang Cheng, Associate Professor of Beijing University of Posts and Telecommunications, and Yao Quanming, Assistant Professor of Tsinghua University.

Associate Professor Huang Gao gave a wonderful explanation about the Transformer infrastructure for long sequences, and conducted an in-depth analysis of the advantages and challenges of the Transformer model in processing long sequence data. Associate Professor Yang Cheng shared the innovative ideas of an efficient collaboration framework for large language model agents and proposed new methods to optimize the efficiency of multi-agent collaboration. Assistant Professor Yao Quanming discussed the simplicity principles and technical implementation in deep learning from the perspective of minimalism, and raised the key issue of how to simplify the model while maintaining high performance.

Embodied Intelligence and Intelligent Interaction Theme Forum

At the Embodied Intelligence and Intelligent Interaction Theme Forum, experts and scholars from different academic institutions discussed embodied intelligence and intelligent interaction. Wonderful speech. The forum was hosted by Li Baiyang, assistant professor of Nanjing University, and invited academic figures such as Sui Yanan, associate professor of Tsinghua University, Zhao Dongyan, professor of Peking University, and Li Hongyang, assistant professor of the University of Hong Kong, to give keynote speeches.

Associate Professor Sui Yanan deeply discussed the self-model of embodied intelligence and analyzed how to improve the autonomy and adaptability of intelligent agents through self-modelling. Professor Zhao Dongyan shared the intelligent question and answer technology based on large-scale language models and elaborated on the application and development of language models in complex question and answer systems. Assistant Professor Li Hongyang brought research on the challenges and opportunities of embodied intelligence in control tasks, raised the "Achilles' Heel" problem, and emphasized the key bottlenecks and future research directions in the control capabilities of intelligent agents.

Trusted AI and Large Model Security Theme Forum

At the Trustworthy AI and Large Model Security Theme Forum, experts and scholars in many fields focused on the credibility and security of artificial intelligence were discussed in depth. The forum was chaired by Professor Wang Ziwei of the Beijing Institute of Graphic Communication. Speakers included Han Bo, Assistant Professor of Hong Kong Baptist University, Qu Liangqiong, Assistant Professor of the University of Hong Kong, and Noor Zaman Jhanjhi, Professor of Taylor's University and other internationally renowned scholars.

Professor Han Bo shared how to explore the construction and application of trustworthy basic models under imperfect data, and proposed how to ensure the credibility and effectiveness of the model when data is incomplete. From the perspective of federated learning, Professor Qu Liangqiong discussed how to promote the application and development of this technology in multi-party collaboration through heterogeneity evaluation, optimization and privacy protection. Prof. Noor Zaman Jhanjhi focused on network security issues in the era of generative AI and analyzed the current generationSecurity challenges existing in conventional AI applications and proposed countermeasures.

Application Innovation Theme Forum

At the Application Innovation Theme Forum, experts shared the latest progress and innovative ideas in the practical application of artificial intelligence technology. The forum was chaired by Li Hongjun, a professor at Beijing Forestry University, and specially invited Lu Lei, an assistant professor at King's College London, and Xia Xin, Huawei's chief expert in software engineering application technology, to give keynote speeches.

Professor Lu Lei discussed in depth the application of deep learning in the field of cardiovascular health, demonstrating how AI can promote innovation in medical diagnosis and improve early detection and treatment of cardiovascular diseases. Expert Xia Xin focused on the application of large models in software engineering and analyzed the progress and challenges brought by large model technology, especially the application potential and technical bottlenecks faced in the development of complex software systems.

On December 22, the conference continued to carry out in-depth academic discussions, and held the "Large Model Technology Exploration and Practice Theme Forum" and the "AI Frontier Application Theme Forum" online and offline in parallel. Industry experts from well-known domestic and foreign universities such as the Chinese Academy of Sciences, Beijing University of Information Science and Technology, Binghamton University, and the University of Reading, as well as companies such as Tencent and Microsoft Health Futures, shared the cutting-edge technology development and interdisciplinary applications of large models. technological breakthrough. These wonderful speeches not only provide a new perspective for the theoretical exploration of large model technology, but also provide an operational path for its practical application. Judging from current development trends, interdisciplinary collaboration and deep integration of technology will become the key driving force for the development of large model applications, promoting the release of AI innovation potential in a wider range of fields.

Large Model Technology Exploration and Practice Theme Forum

At the Large Model Technology Exploration and Practice Theme Forum, experts conducted in-depth discussions on the development trends, applications and challenges of large models. Discuss and share. Song Zhigang, associate researcher at the Institute of Semiconductors, Chinese Academy of Sciences, presided over the forum. Speakers included Xu Jungang, professor at the University of Chinese Academy of Sciences and director of the Cloud Computing and Intelligent Information Processing Laboratory, Wu Xian, head of the Tianyan Research Center of Tencent Youtu Laboratory, and Information Technology of the Chinese Academy of Sciences. Tang Xuehai, a researcher at the Institute of Engineering, Pang Liang, an associate researcher at the Key Laboratory of Intelligent Algorithm Security at the Institute of Computing Technology, Chinese Academy of Sciences, and You Xindong, a professor at Beijing University of Information Science and Technology and other internationally renowned scholars. Professor Xu Jungang made an in-depth analysis of the latest development trends and application prospects of large models. Wu Xian shared the transformation from deep learning to large models, focusing on innovative attempts in medical AI. Researcher Tang Xuehai proposed new ideas for large model security governance. Associate Researcher Pang Liang further clarified the far-reaching significance of large model applications by explaining the cutting-edge technology and social impact of retrieval-enhanced large models; Professor You Xindong discussed the key technologies of knowledge graphs and their practical applications in multiple fields.

AI Frontier Application Theme Forum

At the AI ​​Frontier Application Theme Forum, top experts from different academic and industry fields shared their cutting-edge research results and application explorations. The forum is sponsored by Hong Kong Baptist UniversityProfessor Han Bo hosted the event, and speakers included Binghamton University Professor Zhongfei Zhang, Microsoft Health Future General Manager Hoifung Poon, University of Reading Professor and Computer Science Department Director Shuanghua Yang, and University of Illinois at Chicago Professor Liu Bing. Prof. Zhongfei Zhang discussed the application of uncertainty analysis in out-of-distribution detection. Hoifung Poon, general manager of the future of health at Microsoft, demonstrated the potential of AI in accelerating health development. Shuanghua Yang introduced the importance of multi-view learning in multi-variable time series anomaly detection. Professor Liu Bing analyzed the application of large language models in continuous learning. IGC 2024 —

International Conference Special Session

The afternoon of December 22 is the AIGC 2024 International Conference Special Session, which brings together people from Communication University of China, Dalian University of Technology, Beijing University of Posts and Telecommunications, and the Chinese University of Hong Kong , Delft University of Technology, Yamaguchi University, Russian Academy of Sciences, Kent State University and other more than 20 universities and teachers and students from more than ten countries. Starting from different disciplinary backgrounds, they demonstrated the development trends of AI technology, interdisciplinary applications, as well as the latest breakthroughs and future prospects of large model technology in various fields, providing a wealth of academic perspectives and inspiration.

Offline meeting

Online meeting

Conclusion

The two-day meeting not only provided experts and scholars from all over the world with It is an academic platform for in-depth communication and ideological collision, and is of great significance in promoting the interdisciplinary development of generative AI and large model technology. The successful holding of the conference marks that the application research of generative AI and large model technology will further develop on a global scale. As technology continues to advance, academia and industry will continue to work hand in hand to jointly promote the innovation and application of AIGC in a wider range of fields.

We firmly believe that AIGC is not only an academic forum, but also a platform that promotes global cooperation and promotes the deep integration of artificial intelligence and industry. We look forward to continuing to move forward together at the AIGC 2025 conference and exploring the infinite future of AIGC together!